www.CalSolarResearch.ca.gov Final Project Report: Integration of High Penetration Renewables Using Distributed Energy Resources: A Case Study on the University of California, San Diego Grantee: Viridity Energy June 2013 California Solar Initiative Research, Development, Demonstration and Deployment Program RD&D:
93
Embed
Final Project Report - Californiacalsolarresearch.ca.gov/images/stories/documents/sol2_funded... · Final Project Report: Integration of High Penetration Renewables Using Distributed
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
www.CalSolarResearch.ca.gov
Final Project Report:
Integration of High Penetration
Renewables Using Distributed
Energy Resources: A Case
Study on the University of
California, San Diego
Grantee:
Viridity Energy
June 2013
California Solar Initiative
Research, Development, Demonstration
and Deployment Program RD&D:
PREPARED BY
1801 Market Street, Suite 2701 Philadelphia, PA 19103 484.534.2222
Principal Investigator: Nancy Miller
Project Partners: Energy and Environmental Economics, Inc.
PREPARED FOR
California Public Utilities Commission California Solar Initiative: Research, Development, Demonstration, and Deployment Program
Additional information and links to project related documents can be found at http://www.calsolarresearch.ca.gov/Funded-Projects/
DISCLAIMER
“Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the CPUC, Itron, Inc. or the CSI RD&D Program.”
The goal of the California Solar Initiative (CSI) Research, Development, Demonstration, and Deployment (RD&D) Program is to foster a sustainable and self-supporting customer-sited solar market. To achieve this, the California Legislature authorized the California Public Utilities Commission (CPUC) to allocate $50 million of the CSI budget to an RD&D program. Strategically, the RD&D program seeks to leverage cost-sharing funds from other state, federal and private research entities, and targets activities across these four stages:
Grid integration, storage, and metering: 50-65%
Production technologies: 10-25%
Business development and deployment: 10-20%
Integration of energy efficiency, demand response, and storage with photovoltaics (PV)
There are seven key principles that guide the CSI RD&D Program:
1. Improve the economics of solar technologies by reducing technology costs and increasing system performance;
2. Focus on issues that directly benefit California, and that may not be funded by others;
3. Fill knowledge gaps to enable successful, wide-scale deployment of solar distributed generation technologies;
4. Overcome significant barriers to technology adoption;
5. Take advantage of California’s wealth of data from past, current, and future installations to fulfill the above;
6. Provide bridge funding to help promising solar technologies transition from a pre-commercial state to full commercial viability; and
7. Support efforts to address the integration of distributed solar power into the grid in order to maximize its value to California ratepayers.
For more information about the CSI RD&D Program, please visit the program web site at www.calsolarresearch.ca.gov.
Model campus resources ......................................................................................... 2 Proposed business strategies .................................................................................. 3
Key results and discussion .......................................................................................... 5 Conclusions, recommendations, benefits to California ................................................ 6
Summary of baseline task ...................................................................................... 32 Test and Analyze Scenarios and Publish Results ..................................................... 33
Tariff and incentive insights ....................................................................................... 55 Findings ......................................................................................................................... 57 Recommendations ........................................................................................................ 59 Public benefits to California ........................................................................................... 60 References and Companion Reports ............................................................................ 62
VPower as a What-if Analysis Tool ............................................................................ 72 Example VPower Cases ........................................................................................ 72
the campus and in part due to the team focusing on VPower software modifications and
debugging efforts to support the accurate simulation of campus microgrid operations.
A considerable portion of the project was focused on gathering and analyzing campus
historical data from several system sources in order to validate models and understand
baseline operations under current rates. This scenario is not unusual based on
experiences with other customers, but is frequently underestimated.
Conclusions, recommendations, benefits to California
The feasibility of managing DER in a simulation and real time environment was
evaluated by using the UCSD Dispatch Optimization Tool and the VPower platform to
model and optimize dispatch schedules for micro grids during this project.
The team’s findings using the UCSD Dispatch Optimization Tool suggest that DER are
technically capable of providing cost-effective integration services. These findings,
however, also suggest that incentive and program design changes are needed to
strengthen the business case for large C&I customers.
Peak load shifting. Removing the non-coincident demand charge (which would require
recovering SDGE’s fixed costs elsewhere) could increase a large C&I customer’s cost-
effective peak-load reduction. In the case of UCSD, the estimated increase is about ~ 1
MW.
PV firming. PV firming by a large C&I with its own resources feasible but may be more
costly than relying on the grid. Based on the UCSD case, there is a need for further
research to assess the cost-effectiveness of using alternative cost and tariff structures
applied at the distribution level.
Grid support. A large C&I customer may profitably offer grid support service. For the
UCSD case, small energy cost savings were found. But the savings can increase with
additional resources enlisted to provide independent up or down regulation bids.
P a g e | 7
The above findings lead to the following recommendations to policy makers:
Waive the non-coincident demand charge for PLS customers to promote greater
peak load shifting and increased off-peak load to absorb excess generation.
Allow utilities to negotiate terms on an individual basis with large C&I customers
to accommodate unique capabilities and appropriate, site specific baseline
calculations.
Support development of an operationally robust dispatch model that accounts for
uncertainty and assesses the benefits and risks from complex operational
strategies. Also develop computationally efficient optimization approaches hourly
or sub-hourly dispatch over daily, monthly and annual time steps with more
powerful optimization engines.
Support an implementation study of DER integration strategies using UCSD as a
pilot site. Modest additional effort would leverage this work and use UCSD as a
case study produce a great deal of information on how modeled strategies
translate to real world operation.
The insights from this study are relevant beyond UCSD. There is significant technical
potential for using existing DER at C&I customers across California to provide
renewables integration services. College campuses total 500 MW of load; industrial
customers total over 2000 MW of load1 and have many controllable end-use loads
(pumps, fans, motors); there are ~ 8500 MW of combined and heat and power systems
at ~ 1,200 sites in California2. Many of these customers have similar DER system types
as UCSD and could potentially provide renewables integration services. This analysis
shows that a simple policy change —removing the non-coincident demand charge can
decrease load by ~ 1 MW at UCSD.
This project has generated insights, tools and strategies beyond renewables integration.
In particular, similar analysis can be done for California campuses to reduce their
1 Itron 2007, Assistance in Updating the Energy Efficiency Savings Goals for 2012 and Beyond Task A4 .
1 Final Report : Scenario Analysis to Support Updates to the CPUC Savings Goals Main (2007), at 37. 2 ICF International, 2012. Combined heat and power: Policy analysis and 2011-2013 market assessment.
Report prepared for the California Energy Commission. Report CEC-200-2012-002
P a g e | 8
overall energy consumption, costs and GHG emissions, which is highly relevant in an
era of cost consciousness and university sustainability goals.
P a g e | 9
Introduction
Viridity Energy, Inc. (Viridity) and Energy and Environmental Economics, Inc. (E3)
received a grant under the California Solar Initiative (CSI) Grant Solicitation 2 to study
innovative models, rates and incentives to promote integration of high penetration PV
with real-time management of customer-sited distributed energy resources (DER).3 This
work is motivated by numerous policies promoting renewable and distributed generation
in California. The CSI has a target of 1940 MW of new solar capacity by 2016 in support
of the State of California’s Million Solar Roofs Program and the California Renewable
Portfolio Standard (RPS) that requires 33% of energy procurements by energy suppliers
to be procured from eligible renewable energy resources (including solar resources) by
2020. Numerous studies highlight the potential challenges from high penetration of
intermittent renewable generation.
The University of California, San Diego (UCSD) provided the host site for this project.
The UCSD microgrid has a rich DER base that includes a 2.8 MW fuel cell powered with
directed biogas, a central utilities plant with 30 MW of electrical generation, steam and
electric chillers, a 3.8 million gallon thermal energy storage (TES) and roughly 1.5 MW
of onsite solar PV, including two sites with PV integrated energy storage. UCSD owns
and maintains a 69 kV transmission substation and four 12 kV distribution substations
on campus, with multiple PMU synchrophasors installed by SDG&E. UCSD is also in
the process of installing over 50 Level 2 & 3 electric vehicle charging stations.
The goals of this project were to install Viridity’s VPower platform and demonstrate
dispatch and optimization strategies using UCSD resources to support the integration of
renewable and distributed generation. This project’s approach was to characterize the
campus resources in both the VPower platform, and the UCSD Dispatch Optimization
Tool developed by E3. The models within these applications were used to test the
impacts and cost-effectiveness for three types of strategies: peak load shifting, PV
firming, and grid support. The results show that cost-effective integration strategies are
possible with DER’s and identify specific tariff and market barriers encountered.
3 CSI Solicitation #2 was titled “Improved PV Production Technologies and Innovative Business Models”.
P a g e | 1 0
Project Objectives
The broad goal of this project was to explore how distributed energy resources (DER)
can cost-effectively support the integration of high penetrations of PV systems. The
project develops innovative strategies to accomplish this goal and evaluates these
strategies using the UCSD campus as a case study. The proposed strategies are
designed to overcome current gaps and barriers in energy markets, utility programs and
tariffs.
More specifically, this project focused on the following objectives:
Develop dispatch and optimization strategies for DER to reduce energy costs, integrate renewable generation and support reliable grid operation
Develop tariffs, incentives and business models to promote the adoption of the identified strategies
Enhance the VPower model and smart grid master controller at UCSD in order to test and demonstrate identified strategies with centralized dispatch and optimization of campus DER in real time and simulation modes
Perform cost-benefit analysis for the feasible strategies from societal, utility, customer and ratepayer perspectives
Provide documentation and an analysis tool to disseminate actionable findings to other large commercial and industrial customers and policymakers
The project scope included eight tasks to accomplish these objectives as described in
detail under subsequent sections of this report.
P a g e | 1 1
Project Approach and Results
The project team focused on three key areas: (1) analyze and demonstrate how load
management and shifting can serve as a flexible resource to integrate a high
penetration of PV systems, (2) demonstrate that barriers to the deployment of high
penetration PV systems can be overcome through providing appropriate incentives and
tariff changes coupled with real-time resource monitoring and management and (3)
demonstrate that the optimization of distributed energy resources in support of
increased deployment of PV systems can provide economic, reliability and market price
benefits to California .
The eight tasks identified for the grant project can be broadly categorized as:
Project Management and Reporting
Task 1: Project Management, Reporting, Technology Transfer and Outreach Preparation and Environment Setup
Task 2: Identify and specify strategies for integrating high penetration PV with DER management at UCSD in the California market using VPower™ Task 3: Identify and develop tariffs and incentives that promote the adoption and optimal dispatch of promising DER technologies and load management strategies Task 4: Install and Integrate VPower System (in simulation and real time) Baseline Campus Resource Performance
Task 5: Establish baseline performance for the UCSD DER operation under current rates and incentives Test and Analyze Scenarios and Publish Results
Task 6: Refine and test business models, management strategies, tariff and incentives with VPower in simulation and real-time mode Task 7: Cost Benefit Analysis Task 8: Analysis Tool
P a g e | 1 2
Project Management and Reporting Task 1 focused on project management and dissemination of information through a
variety of reports and presentations to stakeholders. Monthly status reports and bi-
annual reports highlighted the project progress and challenges. The final report (this
report) summarizes the project activities, findings, recommendations and lessons
learned.
Project stakeholders were further informed through an introductory presentation in early
2011 (California Solar Initiative (CSI) Grant Project Overview ) and a mid-project
overview and demonstration of VPower (California Solar Initiative (CSI) Research,
Development, Demonstration and Deployment Program Grant #2 Demonstration) in
early 2012.
Preparation and Environment Setup
Objectives
The main objectives of tasks 2, 3 and 4 were to (1) identify dispatch and optimization
strategies for DER and load management at UCSD to reduce energy costs, support the
integration of renewable solar generation and support reliable grid operation, (2)
identify and develop tariffs and incentives that can encourage cost-effective
incorporation of DER into the California power grid; and (3) enhance the VPower model
and microgrid master controller at UCSD in order to test and demonstrate the identified
strategies with centralized dispatch and optimization of campus DER in real time and
simulation modes.
A further objective was to receive real time and forecast information from the various
mix of metered campus resource types including generation, load, hot and cold water
requirements, price data, and weather data, and to recommend an optimized campus
dispatch. The recommended dispatch was to be reviewed and validated by the campus
operators and eventually used in a real-time dispatch.
The variation of UCSD’s thermal and electrical needs on weekends and from year to
year were investigated.
Overall findings on UCSD’s electrical and thermal needs are as follows.
Thermal and electrical loads exhibit significant base load across all hours with
electrical load factors significantly larger than thermal load factors
Electrical consumption exhibits the least amount of variability; hourly profiles follow
typical shapes with high load factors
Thermal loads exhibit significant variability across months and hours; thermal loads
show strong seasonal dependence
Variability for a specific hour within the month is due both to year-to-year and day-to-
day varying conditions
Variability in loads exists across years but these do not appear to be load growth
related; year to year effects likely driven by temperature variability
P a g e | 2 7
The historical loads provide insight into future loads. The time series data along with
statistical characteristics of the load data can be used to develop dispatch schedules
and assess how sensitive these schedules are to uncertainty in future loads.
Central plant efficiency and daily operations
The loading order, steam utilization and overall efficiency of UCSD’s CHP system were
assessed. The overall efficiency is defined as the ratio of useful output to useful input.
The steam utilization by ‘dispatchable’ variable load systems is defined as the fraction of
steam generated by the natural gas generators that is used by steam chillers, turbine, or
hot water heat exchanger (other ‘base load’ systems use a constant amount of steam
year-round); thus the steam utilization will be less than 100% for the dispatchable
variable load systems. The results of this analysis are shown in Figure 9 and Figure 10.
Figure 9: Steam utilization of combined heat and power system
P a g e | 2 8
Figure 10: Combined heat and power system operation and efficiency
The overall efficiency of UCSD’s CHP system is ~ 60-70% with a median value of 65%;
for context, the CPUC 2010 Impact Evaluation report for the Self Generation Incentive
Program reported total system efficiencies ranging from ~40-65%. The natural gas
generator components are ~ 30% (12,000 Btu/kWh heat rate). The median steam
utilization was estimated as ~ 75%. (Due to improvements in base load steam utilization
systems, future utilization of ~ 85% is expected.)
The loading order can be observed in these figures. The steam chillers are fueled with
steam, followed by the hot water systems, and lastly by the steam turbine. This
empirical loading order is largely consistent with UCSD’s heuristics, although in the
winter, hot water generation may be favored over chilled water generation.
Figure 11 shows an example of daily onsite generation, solar production, thermal
needs, chiller operation and thermal storage operation using data from June 7, 2011.
The generators typically operate at full capacity and in steady state mode; they are
generally operated no lower than 10 MW (or 77%). The steam chillers are operated at
constant rate across most hours. Electric chiller output is avoided during peak hours
(11-6 pm) and greatest when charging the thermal energy storage (TES) tank. The TES
tank is charged at night until the early morning and discharged during peak hours. The
P a g e | 2 9
TES tank is operated conservatively such that it maintains capacity to compensate for
unexpected steam chiller outages; that is, the TES tank is not discharged at the
‘optimal’ rate assuming perfect foresight on chilled water demand. For the example
shown, solar PV output is ~ 1 MW or 3% of the electrical load.
Figure 11: Daily operations, June 7 2011 example
Minimizing campus energy costs is a complex process that involves optimizing the CUP
generation for optimal use of recovered energy and operating the TES tank and electric
chillers to minimize energy and demand charges. The presence of an all hours demand
charge complicates the operations because turning on the electrical chillers and turning
off the generators during off-peak periods risks moving the maximum demand, which
determines the all-hours demand charge, to the off-peak period.
Efficiency and output of individual systems
The efficiencies and operating capacities of the following systems were characterized:
the natural gas and steam generators, chillers, boilers, and thermal storage tank. An
example is shown in Figure 12 which shows the efficiency curve for an electric chiller.
P a g e | 3 0
Figure 12: Example of chiller efficiency, electric chiller #5
The salient points from the analysis of the CUP’s individual systems are summarized as
follows:
UCSD’s electric chillers are centrifugal and their efficiencies ranges from ~ 0.5 to 0.7
kW/ton (coefficient of performance, COP ~ 5-6).
The steam driven chiller efficiency ranges from ~ 8000 Btu/hr/ton to 10,500 Btu/hr
/ton (COP ~ 1.4-1.7)5. The chillers’ operating capacities are generally ~ 70-90%
(50th & 90th percentiles) of nameplate capacities.6
The two natural gas generators have a median heat rate of ~ 11.5 MMBtu/MWh &
have an output of ~ 98% of nameplate capacity; steam generator is ~ 15% efficient
Three boilers have median efficiency of ~ 75%
The TES tank has a median daily discharge of 16,250 ton-hours; overall losses of ~
4.4%; median hourly discharge rate of ~ 1330 ton or 15 MMBtu/hr (capacity of ~
3100 ton or 35 MMBtu/hr).
Figure 13 shows the solar output of UCSD solar PV system by month for 2011. The
winter output peaked at ~ 500 KW and summer output at ~ 850 KW.
5 The relative COPs Electric chillers are known to have greater efficiencies than condensing steam-
turbine driven chillers; the efficiencies observed here are within range of that expectation. 6 Exceptions include steam chiller WC1 ~ 40-50% and electric chillers WC 7 & 9 ~ 50-80%. WC1
underwent a refrigerant retrofit, which accounts for its low loading.
P a g e | 3 1
Figure 13: Average hourly solar output by month
Preliminary business analysis
Using the system efficiencies and capacities identified, a simple dispatch model using
Excel was developed to compare the costs of four basic operating scenarios:
UCSD imports its electrical needs fully
UCSD imports its electrical needs and operates the TES tank
UCSD operates its onsite generation
UCSD operates its onsite generation and TES tank
This model was limited in that ramp rates, minimum run times were not explored. The
generator schedules were also not varied from day to day but was assumed to be
constant throughout the year. The model also assumes a loading order of the steam
utilizing systems and operation of the TES tank, rather than solving for the optimal
dispatch solutions.
P a g e | 3 2
Table 1: Estimated Energy Costs between June 1 and October 31, 2011
Table 1 shows the results of the preliminary business analysis. This analysis illustrates
that UCSD is motivated economically to operate its cogeneration (CHP) system and
TES tank. This finding is important because these systems are anticipated to be
important for providing renewables integration services.
Summary of baseline task
The baseline task provided the following information:
Quantitative inputs for the business models, such as individual and system level
across all cases. The model does not optimize any of the systems outside of the CUP,
such as other HVAC systems, backup generators or auxiliary equipment. Focusing on
one year of historical data, the model considers changes in variable operating costs
only. The fixed cost of existing equipment is considered sunk, and no capital investment
in new resources on campus is contemplated.
The model incorporates operating constraints in the form of upper and lower operating
capacities, startup costs, and minimum run times for CUP equipment. The optimization
engine must determine schedules that meet UCSD’s electrical and thermal
requirements while satisfying these operating constraints. A key feature of the model is
TES tank management: the model determines discharge and charge schedules subject
to charge/discharge rates, such that monthly demand charge is minimized.
The model optimizes over two different time frames, minimizing for either daily total cost
or monthly total cost. The monthly approach was important for capturing demand
charges accurately, which requires knowledge of demand over the entire month.
Because the optimization model has perfect foresight over the period being solved (for
example, the electrical demand 12 hours away), using two different time frames allowed
for a balance between more or less forward looking results.
For the monthly minimization, due to computational limits, the temporal and resource
resolution was reduced. Rather than developing hourly schedules, bi-hourly schedules
were developed; chillers were aggregated and minimum run times were not imposed on
these systems. Bi-hourly campus demands were generated from hourly data.
The daily minimization, which runs for consecutive days, requires constraints to be
satisfied each hour of the day and passes the operating state of each resource (e.g.
whether a resource is on, and for how many hours it has been running) and maximum
demand level from one day to the next. The daily time frame affords greater time
resolution at the expense of suboptimal results for the demand charge and TES tank
management. The monthly time frame does not offer the same temporal granularity, but
produces optimal solutions for demand charges. These two approaches can be
integrated by feeding month long optimization results into the daily optimization.
P a g e | 4 4
Figure 17: Screen shot of UCSD Dispatch Optimization Tool
The following approximations were made to reduce solving time and to maintain the
number of variables and constraints within the software limits:
Chiller, generator and boiler efficiencies are represented as constant values,
rather than a function of the output
Steam production is assumed to be constant from the natural gas generators
between minimum and maximum electrical output operating levels; this
assumption is based on UCSD’s operational experiences
The boilers are represented as an aggregated single unit rather than as three
independent units
The TES must be fully recharged at the end of the period of total cost
minimization
For the month long model additional approximations are required:
A bi-hourly time step, with hourly campus needs and prices averaged over every
two hours into a single time steps
Only gas turbine minimum run times are included
P a g e | 4 5
Individual chillers are aggregated into composite chillers, steam and electric, with
weighted average efficiencies
Strategy and Software Functionality Comparison
As the project team evaluated the target strategies, it was clear that additional VPower
software enhancements were required. Without the real time data feed, the capability to
perform studies over longer periods with large amounts of varying inputs and to graph
results was limited. Viridity decided to streamline the development effort through their
core software development team rather than the project team.
Ultimately, this resulted in two actions which supported meeting project objectives.
First, the UCSD Dispatch Optimization Tool would be leveraged to perform the bulk of
analysis until the relevant capabilities were available in VPower. Second, Viridity
VPower core product development team priorities were established to produce the
needed system enhancements.
The VPower release made available in April 2013, has the capabilities to optimize and
analyze opportunities on a longer term basis. However, limited time was available
within the grant schedule to fully develop and execute simulations leveraging the new
capabilities and include those results in this report. As a result, the details of the
enhancements and some initial results are included in Appendix A.
Strategy Analysis
The results from the UCSD Dispatch Optimization Tool are summarized below for each
strategy category in three ways.
First, an example week of resource dispatch, showing how the dispatch of
campus resources changes with each successive case. This illustrates how
the strategy impacts the dispatch of campus resources, and how changing
constraints or available resources alters that dispatch.
Second, the change in net cost from the base case for each type of cost for
the campus is shown: electricity import costs, demand charges, natural gas
costs, incremental revenues (if any) and the net impact of all four summed
together.
P a g e | 4 6
Finally, the net cost impact for a summer and winter month is shown.
Although the optimization for each case was performed over the entire one year period
of analysis (June 2011 – July 2012), a subset of results is presented below for two
reasons. First, it is far easier to effectively represent and highlight impacts over the
weekly or monthly time frame than it is for a full year of hourly data. Second, due to
computational limitations, the model did not solve consistently for all the days and
months of the year. In all cases, more that 93% of the days/hours solved in the
optimization, giving a good representation of performance across the year and varying
conditions. A winter and a summer month were chosen that were directionally
consistent with the results for most months of the year.
Examples of the hourly dispatch results in the month of August that illustrate the value
of utilizing different levels of UCSD DERs are shown in Figure 18, with the three graphs
showing the full import, full import with TES and cogeneration with TES dispatches.
These three graphs show how the addition of resources changes the campus dispatch
due to lower costs.
The figure shows that campus electrical load is met by imports alone, and the electrical
chillers run consistently throughout the day except for a few hours with high energy
prices. In these few hours the steam chillers provide cooling, effectively fuel switching to
natural gas.
The middle panel shows how the dispatch changes with the addition of the TES system.
The electrical chillers turn off during some afternoons where the model discharges
stored chilled water to avoid increasing the on-peak demand rate and high priced
energy.
The final panel shows the cogeneration with TES case. The cogeneration substantially
decreases the level of imports, and its steam to drive chillers, together with the TES,
means the electric chillers are not needed.
P a g e | 4 7
0
10
20
30
40M
W
Full Imports Energy Dispatch
0
10
20
30
40
MW
Full Imports with TES Energy Dispatch
0
10
20
30
40
8/6
h0
8
8/6
h1
6
8/7
h0
0
8/7
h0
8
8/7
h1
6
8/8
h0
0
8/8
h0
8
8/8
h1
6
8/9
h0
0
8/9
h0
8
8/9
h1
6
8/1
0 h
00
8/1
0 h
08
8/1
0 h
16
8/1
1 h
00
8/1
1 h
08
8/1
1 h
16
8/1
2 h
00
8/1
2 h
08
8/1
2 h
16
8/1
3 h
00
8/1
3 h
08
8/1
3 h
16
8/1
4 h
00
8/1
4 h
08
8/1
4 h
16
8/1
5 h
00
8/1
5 h
08
8/1
5 h
16
8/1
6 h
00
MW
Cogeneration with TES Energy Dispatch
NG Turbine 1 Production MWh NG Turbine 2 Production MWh Steam Turbine Production MWh
Imports to Campus MWh Electric Chiller Consumption MWh Campus Electric Demand MWh
Figure 18: Examples of hourly dispatch for the stepwise analysis, where cases have progressively more DERs in each graph
Peak Shifting Strategy: Removing all hours demand charge
Analysis of historical campus loads and resources showed that the all-hours demand
charge frequently limits off-peak charging of the TES tank. In some cases, fully
recharging the TES during off-peak hours would cause an increase to the maximum
demand billing determinant for the month; that is the UCSD off-peak demand would
exceed their previously set on-peak demand MW for the month. This leads to a counter-
P a g e | 4 8
productive result for a customer with load-shifting capability wherein UCSD is prevented
from reducing peak loads to the full extent possible.
Peak Shifting Strategy: Reducing duration of peak window
The shorter summer peak period strategy results in a minor change in dispatch as
compared to the base case as shown in Figure 19. While imports consistently remain
low over all hours for both the shorter peak and base case, the gas turbines are
dispatched at a marginally lower level in some hours in the shorter peak strategy.
0
5
10
15
20
25
30
35
40
MW
Base Case Energy Dispatch
0
5
10
15
20
25
30
35
40
MW
Shorter Summer Peak Window Energy Dispatch
0
5
10
15
20
25
30
35
40
8/8
h0
1
8/8
h0
9
8/8
h1
7
8/9
h0
1
8/9
h0
9
8/9
h1
7
8/1
0 h
01
8/1
0 h
09
8/1
0 h
17
8/1
1 h
01
8/1
1 h
09
8/1
1 h
17
8/1
2 h
01
8/1
2 h
09
8/1
2 h
17
8/1
3 h
01
8/1
3 h
09
8/1
3 h
17
8/1
4 h
01
8/1
4 h
09
8/1
4 h
17
8/1
5 h
01
8/1
5 h
09
8/1
5 h
17
8/1
6 h
01
8/1
6 h
09
8/1
6 h
17
8/1
7 h
01
8/1
7 h
09
8/1
7 h
17
8/1
8 h
01
8/1
8 h
09
8/1
8 h
17
8/1
9 h
01
8/1
9 h
09
8/1
9 h
17
MW
No All Hrs Charge, Higher On-Peak Demand Rate Energy Dispatch
NG Turbine 1 Production MWh NG Turbine 2 Production MWh Steam Turbine Production MWh
Imports to Campus MWh Electric Chiller Power Consumption MWh Campus Electric Demand MWh Figure 19: Examples of hourly dispatch for the Peak Load Shifting
P a g e | 4 9
PV Firming Strategy: Handle error with Gas Turbines
The natural gas generation firming strategy shows savings in electricity imports relative
to the base case, but those savings are overwhelmed by increased natural gas costs
making total costs increase in both months.
PV Firming Strategy: Two-part rate
All the PV firming strategies have positive net costs but with varying levels. The 2 part
tariff strategy incurs about 1% higher electricity import costs, which include penalty
payments and negligible increases in natural gas costs for both months. The increase in
electricity imports is higher for the grid leaning strategy than the other strategies (Figure
20).
P a g e | 5 0
0
5
10
15
20
25
30
35M
WPV Firming Leaning on Grid Energy Dispatch
0
5
10
15
20
25
30
35
MW
PV Firming NG Gen Energy Dispatch
0
5
10
15
20
25
30
35
8/1
1 h
08
8/1
1 h
14
8/1
1 h
20
8/1
2 h
02
8/1
2 h
08
8/1
2 h
14
8/1
2 h
20
8/1
3 h
02
8/1
3 h
08
8/1
3 h
14
8/1
3 h
20
8/1
4 h
02
8/1
4 h
08
8/1
4 h
14
8/1
4 h
20
8/1
5 h
02
8/1
5 h
08
8/1
5 h
14
8/1
5 h
20
8/1
6 h
02
8/1
6 h
08
8/1
6 h
14
8/1
6 h
20
8/1
7 h
02
8/1
7 h
08
8/1
7 h
14
8/1
7 h
20
8/1
8 h
02
8/1
8 h
08
8/1
8 h
14
8/1
8 h
20
8/1
9 h
02
MW
PV Firming NG Gen & Steam Turbine Energy Dispatch
NG Turbine 1 Production MWh NG Turbine 2 Production MWh Steam Turbine Production MWh
Imports to Campus MWh Electric Chiller Consumption MWh Campus electric demand MWh
Figure 20: Examples of energy dispatch for three different formulations for firming campus PV
Incentives for Grid Support Strategy: Market Products
The dispatch of campus resources three regulation cases (natural gas generators bid
the same up and down MWs; gas generators bid separate up and down MWs; and all
resources bid separate up and down MWs) are shown in Figure 21. The dispatch is
similar in all cases, with some additional imports for electric chiller consumption in the
last case. Although the quantity of regulation offered in each case changes, the dispatch
of campus resources does not. With increasing flexibility in market rules and the
resources offering regulation, the optimizer takes further advantage of the opportunity to
earn revenues in the regulation market (Figure 22), but does not alter the dispatch of
campus resources to do so. There is a dramatic difference in costs and revenues
P a g e | 5 1
between the two cases that require the same quanity to be bid in both directions, and
the two cases that allow different quantities in the up and down direction. Offering the
same quantity in both directions requires the generator to operate near the mid-point of
23.3 MWs to provide regulation. Under normal operation, the generators will operate
predominately at 20 or 26. 6 MWs, or one generator will shut down entirely. The
optimizer is generally choosing to offer regulation when the generators would otherwise
operate at 20 MW. Therefore, the overall level of generation is increased, reducing
imports and increasing natural gas consumption.
0
10
20
30
40
MW
Simple Regulation Energy Dispatch
0
10
20
30
40
MW
NG Turbine Regulation Energy Dispatch
0
10
20
30
40
8/2
h0
88
/2 h
12
8/2
h1
68
/2 h
20
8/3
h0
08
/3 h
04
8/3
h0
88
/3 h
12
8/3
h1
68
/3 h
20
8/4
h0
08
/4 h
04
8/4
h0
88
/4 h
12
8/4
h1
68
/4 h
20
8/5
h0
08
/5 h
04
8/5
h0
88
/5 h
12
8/5
h1
68
/5 h
20
8/6
h0
08
/6 h
04
8/6
h0
88
/6 h
12
8/6
h1
68
/6 h
20
8/7
h0
08
/7 h
04
8/7
h0
88
/7 h
12
8/7
h1
68
/7 h
20
8/8
h0
08
/8 h
04
8/8
h0
88
/8 h
12
8/8
h1
68
/8 h
20
8/9
h0
08
/9 h
04
8/9
h0
88
/9 h
12
8/9
h1
68
/9 h
20
8/1
0 h
00
8/1
0 h
04
MW
Whole Campus Regulation Energy Dispatch
NG Turbine 1 Production MWh NG Turbine 2 Production MWh Steam Turbine Production MWh
Imports to Campus MWh Electric Chiller Consumption MWh Campus Electric Demand MWh
Figure 21: Examples of energy dispatch for different regulation bidding strategies
P a g e | 5 2
-10
-5
0
5
10B
id M
W
Simple Regulation Bids
Bids for Reg Down
Bids for Reg Up
-10
-5
0
5
10
Bid
MW
NG Turbine Regulation Bids
Bids for Reg Down
Bids for Reg Up
-10
-5
0
5
10
8/2
h0
8
8/2
h1
2
8/2
h1
6
8/2
h2
0
8/3
h0
0
8/3
h0
4
8/3
h0
8
8/3
h1
2
8/3
h1
6
8/3
h2
0
8/4
h0
0
8/4
h0
4
8/4
h0
8
8/4
h1
2
8/4
h1
6
8/4
h2
0
8/5
h0
0
8/5
h0
4
8/5
h0
8
8/5
h1
2
8/5
h1
6
8/5
h2
0
8/6
h0
0
8/6
h0
4
8/6
h0
8
8/6
h1
2
8/6
h1
6
8/6
h2
0
8/7
h0
0
8/7
h0
4
8/7
h0
8
8/7
h1
2
8/7
h1
6
8/7
h2
0
8/8
h0
0
8/8
h0
4
8/8
h0
8
8/8
h1
2
8/8
h1
6
8/8
h2
0
8/9
h0
0
8/9
h0
4
8/9
h0
8
8/9
h1
2
8/9
h1
6
8/9
h2
0
8/1
0 h
00
8/1
0 h
04
Bid
MW
Whole Campus Regulation Bids
Natural Gas Turbine Bids MW Steam Turbine Bids MW Electric Chiller Bids MW
Bids for Reg Down
Bids for Reg Up
Figure 22: Regulation bids, up and down, for three different strategies for providing regulation
Public Cost Benefit Tool
For reasons described earlier, the UCSD Dispatch Optimization Tool was developed
beyond the scope initially envisioned for this project to perform detailed dispatch
optimization of UCSD resources over an entire year. This significant expansion of the
functionality of the UCSD Dispatch Optimization Tool consumed significant additional
time and budget. Because resources were diverted to fully developing the optimization
approach, and because the model became significantly more complex as a result, a
simplified public interface for the tool as envisioned in Task 8 was not developed.
P a g e | 5 3
Conclusion
Several conclusions can be drawn from this project. They are presented here in three
categories: Modeling, Operations, and Tariff and incentives.
Modeling insights
Modeling insights come from the efforts to implement VPower and the UCSD Dispatch
Optimization tools and working with UCSD operators to parameterize the cost benefit
optimization model. Working closely with the UCSD energy manager proved
instrumental in validating modeling results and identifying where focused detail is
needed and where reasonable approximations can be made.
Modeling and testing a microgrid for either simulation or real time mode is
complex and requires the availability of subject matter experts: The complexity of
modeling and testing the operational parameters for the variety of resources and their
interactions within a microgrid is a detailed and time-consuming task. Having subject
matter experts that are dedicated as part of the project team is key to working through
challenges. UCSD staff was extremely responsive and helpful, but given their other
responsibilities could not support the time required to work with the team to address
some of the data and model issues encountered.
Integrating thermal resources in optimization is required for robust results: Good
integration of thermal resources and their interactions with other resources in
optimization proved crucial to winning operator confidence in the results. Integration
studies tend to focus on electrical impacts, but heating and cooling are key additional
primary end-uses. The two threshold issues for campus operators are: 1) are the results
credible and intuitive and 2) do they include the downstream impacts in hot and cold
water production.
Separate approaches are needed for monthly and daily period of analysis for
demand charge costs: As is true for many large C&I customers, the monthly demand
charge is a large cost driver for UCSD. Performing a full optimization over one month
P a g e | 5 4
was not feasible in VPower nor the UCSD Campus Dispatch Optimization Tool model
due to computational limitations. It was found to be most expedient to adopt the two
stage approach presented in this study: a one month optimization, with approximations
as needed for computational efficiency, to determine maximum demand for demand
charges and TES dispatch; and a more detailed optimization over one to several days
at a time to perform hourly or sub-hourly dispatch optimization.
Operational insights
Modeling efforts and insights produced results that offer some useful observations on
UCSD resource operation. The scenarios are modeled results and as such they do not
fully capture the detailed considerations and uncertainties faced by UCSD microgrid
operators. However modeling hourly dispatch for a full year has offered insights into
how the strategies examined here could work with actual campus operations.
Integrated optimization and dispatch of campus resources can reduce costs
while providing flexibility: Modeling optimal dispatch of campus resources proves
effective in identifying strategies that can reduce costs or increase flexibility relative to
standard operation. Currently, UCSD applies heuristics to dispatch resources, which are
operated in a pseudo-steady state manner. Characterizing and optimizing campus
resources demonstrates the capacity to perform additional services while meeting
campus demands and achieve additional cost savings.
Incorporating additional resources in dispatch strategies does meaningfully
reduce costs or increase flexibility: In both the PV firming strategies and grid support
strategies, adding resources such as the steam generators or electric chillers to the
available portfolio reduces the comparative campus costs and increases the quantity of
service provided.
PV firming with campus resources appears feasible, but more expensive than
current estimates of grid renewable integration costs: Using renewable integration
cost estimates of $8/MWh generated or $31/MWh of forecast error – on the high end of
renewable integration cost estimates – it was determined that using the campus
P a g e | 5 5
resources to firm PV is not cost-effective. This follows the generally accepted wisdom
that a diverse portfolio of resources over a wider geographic area will be more efficient
in managing variability. Including additional campus resources (such as building loads
or electric chillers) could reduce the campus costs. Furthermore, to the extent that there
are higher local integration costs, DER’s could still prove an economic resource for
renewable integration.
Current prices for regulation are cost-effective for campus but revenues are small
compared to total costs: Campus resources can provide frequency regulation in the
CAISO market at today’s prices cost-effectively. However, net revenues are only ~2% of
the total campus energy cost. Regulation revenue can help justify investments in new
resources, but will be supplemental rather than a main driver of the decision. Because
regulation can be a demanding service with increased risk and O&M costs, additional
incentives or alternative strategies (such as pooled provision of regulation by
aggregated networks of distributed resources) will be necessary to encourage wider
adoption.
Tariff and incentive insights
Operational insights often arose together with insights about how changes in the cost
UCSD faced or the addition of incentives could have substantial positive impacts on the
integration strategies. Modeling shows the strategies in this work can be operationally
possible and further work may show they are operationally feasible, but tariffs and
incentives will be the final determinant of whether these integration strategies can be
deployed.
Off-peak demand charge significantly constrains on-peak dispatch of campus
resources: The SDG&E all-hours demand charge proves to be a significant constraint
to the peak load shifting dispatch for UCSD. Because UCSD has significant load shifting
capacity relative to peak net loads, load shifting frequently increases monthly peak
demand, though it occurs in the off-peak period. Simply implementing alternative tariffs
for recovering fixed costs could increase the peak load shifted by over 1 MW. While the
all-hours demand charge was not modeled with the other strategies, it is expected that it
P a g e | 5 6
will also prove to be a disincentive many strategies for using DER for renewable
integration.
Two-part rates will be needed to encourage DER provision of renewable
integration services: Retail tariffs are relatively blunt instruments and impose
significant risks and potential costs for customers seeking to provide renewable
integration services. It is unrealistic to expect dynamic rates alone to provide sufficient
incentives. In fact, as is seen in the PLS strategies, time differentiated rates can lead to
counter-productive incentives when it comes to renewable integration. Supplemental
tariffs and incentives that can be layered on top of retail rates without compromising
utility fixed cost recovery will be necessary to engage the full potential of DER’s for
renewable integration.
Direct participation in wholesale markets do not provide sufficient incentives for
campus provision of integration or ancillary services: Campuses like UCSD have a
diverse and large portfolio of resources, but emissions, economic and end-use
considerations limit the relative quantity of capacity available for providing grid support.
These services can be cost effective from the grid perspective, but participation results
in revenue that is a small percentage of total campus costs. Additional research or
product development is needed to develop strategies to effectively engage to large C&I
customer DERs in wholesale markets.
P a g e | 5 7
Findings
Implementing optimization tools at UCSD revealed unexpected challenges. The
operation of sophisticated, multi-resource combined energy and thermal systems like
the UCSD Microgrid is extremely complex. An experienced team approached this
project with no illusions about the modeling, optimization and system integration
challenges entailed. Even so, several issues prevented the implementation and
operation of VPower as planned, and forced the team to develop an alternative
approach. Compared to most campuses, acquiring, processing and cleaning data from
multiple sources proved time consuming, even for a well metered campus. The historian
and telemetry need for real-time and near real-time campus data were still being
configured during the course of the project. Static data models supported VPower
modeling, but the nature of the interfaces did not easily support strategy testing. The
team determined that evaluating scenarios would require longer duration and more
flexible analyses and therefore, an alternative approach. Separate, computationally
efficient optimization approaches, were developed for hourly dispatch over daily and
monthly/annual periods. Neither tool fully accounts for uncertainty or increased
operational costs or risks from complex operational strategies.
Security requirements need to be considered and addressed as part of
implementing the management and optimization tools. There were several cases
where changes in campus cyber security policies affected the project. The VPower
installation was removed a few times and needed to be reinstalled and re-authenticated.
Security restrictions at some of the off campus sites that were considered for
participation in the project, limited the team’s ability to gather data needed for modeling.
Security could also restrict the external monitoring and control functions necessary for
rapid customer responses to pricing and dispatch signals.
Value and cost estimates for local, distribution grid support and integration
services are needed, but not readily available. There is little, if any, public cost
estimates for local and distribution level impacts, which are frequently the primary
P a g e | 5 8
limiting concerns for utility operators when it comes to high PV penetration and EV
charging. These services are potentially more valuable and lucrative than wholesale
grid markets. Identifying, developing and monetizing high value services for local grid
support is crucial for increased customer, vendor and service provider engagement.
A public and transparent framework to explicitly compare central, distributed,
load and market based renewable integration and GHG reduction strategies is
needed. Although several initiatives and proceedings are examining long-term planning
and procurement for flexible resources and renewable integration, there remains no
framework to readily evaluate and compare the diverse portfolio of alternative strategies
available to utilities and policy makers. A guiding framework for evaluating the relative
costs and benefits of resources like CTs, energy storage, demand response and the
CAISO Flexi-ramp product in meeting identified system needs would be instrumental in
identifying and developing high value, low cost strategies in each category.
The limited value of net AS market revenues relative to total energy costs
reinforces the importance of non-price strategies to engage the substantial
resources of large C&I customers for integration and ancillary services. In eastern
ISO markets, DER’s now provide up to 10% of the total MW’s enrolled in centralized
capacity markets. Participation in reserve and AS markets is much more limited. This
project’s analysis suggests that access to wholesale markets alone is insufficient to
motivate participation by UCSD and by proxy, other large C&I customers. These
findings, together with the experience in eastern ISO markets, suggests that customer
engagement and outreach will be important elements in encouraging DER to provide
renewable integration.
P a g e | 5 9
Recommendations
Support an implementation study of DER integration strategies using UCSD as a
pilot site: To enable the large existing pool of DER to engage in strategies to enable
greater renewable integration the work that has been done for UCSD will need to
adapted to range of applications and disseminated. While this work models the dispatch
of UCSD resources under proposed renewable integration strategies a vital next step in
realizing these strategies is piloting their actually operation at the campus. The
modeling conducted in this analysis does not address the uncertainty and nuances
facing by system operators. An effort to operationalize these strategies for UCSD would
leverage this work and produce a great deal of information on how modeled strategies
translate to real world operation.
Restructure all-hours demand charge for PLS customers: The current all-hours
demand charge paradoxically reduces the incentive for UCSD to shift load and increase
off-peak generation at a time when system operators are claiming an increased need for
both renewable integration and to replace local capacity lost due to the San Onofre
Nuclear Generating Station outage. Restructuring the all-hours demand charge together
with the on-peak demand charge for UCSD and other customers with significant load
shifting capacity could meet both objectives at little or no cost to utilities or ratepayers.
Allow utilities to negotiate terms specific to individual, large C&I customers:
UCSD is an example of a large, underutilized resource for SDG&E. The all-hours
demand charge is counter-productively limiting peak load shifting, and established
baseline rules base on 10 historical days are too inaccurate and risky for UCSD to enroll
in established DR programs. There is established precedent for utilities to negotiate
special rates for customers considering bypass. A similar policy of allowing utilities to
negotiate customized terms to facilitate the maximum participation by local distributed
resources should be considered.
P a g e | 6 0
Public benefits to California
The results of this project are relevant beyond UCSD and could promote DER adoption
and the use of DER for renewables integration. Although the project did not meet the
original goal of demonstrating specific strategies at UCSD in a live environment, the
results provide useful insights for customers and policy makers that can provide
economic and environmental benefits in the near-term.
Technical potential. C&I loads have significant technical potential to provide
renewables integration strategies in California. College campuses total 500 MW of load;
industrial customers total over 2000 MW of load7 and have many controllable end-use
loads (pumps, fans, motors); there are ~ 8500 MW of combined and heat and power
systems at ~ 1,200 sites in California8.
Simple policy changes. Analysis during this project shows that a simple policy change
— removing the all-hours demand charge can decrease load by ~ 1 MW at UCSD. The
value of reducing load by 10 MW (2% of California campus load) is ~$1.0 Million/year
using 2013 avoided capacity costs. (Capacity value in Local Capacity Requirement
(LCR) area such as San Diego are not publicly available but generally estimated to be
much higher.)
Integration at the distribution level. Analysis during this project suggests UCSD can
firm its solar PV using its own resources at a cost comparable to relying on the grid,
even using relatively high estimates of renewables integration costs. However, local
integration costs are uncertain and could be higher than average integration costs,
which increases the value of using DER to provide firming. The two-part that is
described when firming with the grid can be implemented with smart meters.
Insights on grid support. Analysis during this project of grid support suggests it is
economical for UCSD to provide grid support based on regulation prices but the net
7 Itron 2007, Assistance in Updating the Energy Efficiency Savings Goals for 2012 and Beyond Task A4 .
1 Final Report : Scenario Analysis to Support Updates to the CPUC Savings Goals Main (2007), at 37. 8 ICF International, 2012. Combined heat and power: Policy analysis and 2011-2013 market assessment. Report prepared for the
California Energy Commission. Report CEC-200-2012-002
P a g e | 6 1
benefit to UCSD is relatively low. Ancillary service revenue alone may be insufficient for
motivating loads to provide grid support and alternate products and incentives may be
required.
Beyond renewables integration, this project provides insights, tools and strategies that
can be used by California colleges to support efforts in reducing energy consumption,
costs and GHG emissions. For example, achieving the GHG emissions reductions
called for in the University of California’s Policy on Sustainable Energy Practices (which
encourages carbon neutrality as soon as possible) presents numerous challenges and
will require new analysis tools and innovative strategies such as those described in this
study.
P a g e | 6 2
References and Companion Reports
References
CAISO. (2012). Flexible Ramping Products: Draft Final Proposal. Folsom, California:
CAISO.
Itron (2007). Assistance in Updating the Energy Efficiency Savings Goals for 2012 and
Beyond Task A4.1 Final Report : Scenario Analysis to Support Updates to the CPUC
Savings Goals.
PG&E, SCE, SDG&E. Joint IOU Supporting Testimony. July 1 2011, CPUC Long Term